The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing specific shell commands through the ae-cli utility. The instructions strictly bind these executions to well-defined reference files (e.g., adhoc_run.md, asset_search.md), ensuring that the agent remains within the authorized functional scope for data analysis and governance tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input to derive intent and processes external data such as SQL definitions and report snapshots. This surface is mitigated by explicit instructions to confirm business mappings before execution, the requirement to use JSON string literals for CLI flags, and rigorous boundaries for interpreting dynamic SQL placeholders as defined in references/ai_models/sql.md.\n- [SAFE]: The skill follows standard operational practices for its environment, including the maintenance of a telemetry file (.ae/cli-agent-context.json) and the use of a local utility script (scripts/read_results.py) for processing results. No unauthorized external network operations, hardcoded credentials, or obfuscated payloads were detected.

ae-analysis